Cue constraints and their interactions in phonological perception and production

Abstract

Phonological theory requires at least five representationsMultiple levels of representation such as those in Figure 1 are common in psycholinguistic models of speech production (e.g.Levelt 1989) and comprehension (e.g.McClelland and Elman 1986, McQueen andCutler 1997), but are less often seen in phonological theory (I will discuss the exceptions later on).As in my earlier work, I argue in this paper that phonological theory would profit from going beyond its usual two levels, especially by taking the phonetics seriously.Figure 1, then, contains the minimal number of phonological representations that phonologists can be comfortable working with: the Underlying Form (often called the input in the OT literature) and the Surface Form (often called the output in the OT literature).The underlying form is the discrete representation of the phonological structure of morphemes in the language user's mental lexicon; the Morpheme (or Lemma) mediates in connecting the phonological underlying form to semantic features in the lexicon (Saussure 1916), which again are probably connected to the meaning of the utterance and from there to the pragmatic context (these are not shown in Figure 1, since they are likely to concern phonological theory to a lesser extent).The surface form is the discrete representation of the phonological surface structure and consists of prosodic elements (feet, syllables, segments) and phonological substance (features, autosegments).For most phonologists, the surface form does not contain any concrete continuous phonetic detail, and this is something I agree with; it means that one can do insightful investigations in many areas of phonology by just considering the two discrete phonological representations.As for phonetic processing, this is usually regarded as something that comes 'after' the phonology (e.g.Pierrehumbert 1987, Keating 1985, Gussenhoven 2004).Phonologists tend to argue that the phonetics is therefore not really relevant for autonomous phonological processing (e.g.Hale and Reiss 1998), or that it might be relevant but that its modelling is not a priority for phonological theory (e.g.Hayes 1999: fn. 7).On the basis of the abundant existence of seemingly phonetically-inspired processes in segmental phonology, some phonologists have nevertheless tried to include phonetic considerations of articulatory effort and auditory contrast into the usual twolevel model of phonology consisting only of underlying and surface form; by doing so, one must either propose that the phonological surface structure somehow includes continuous phonetic detail (Jun 1995, Kirchner 1998; cf.Flemming 1995 for a model without underlying forms but with a phonetically detailed surface form) or that discrete phonological processing is somehow sensitive to extralinguistic information on phonetic detail (Steriade 1995(Steriade , 2001)).Following Boersma (1998), I take the third possible stand, which takes seriously the possible relevance of phonetics for phonological theory without sacrificing the representational modularity of the phonology and the phonetics: in Figure 1, therefore, the phonetic representations are separate from the phonological representations, but are taken just as seriously.The minimum number of phonetic representations that phoneticians can be comfortable working with are two: the Auditory Form and the Articulatory Form.The auditory form is the continuous representation of sound; it consists of noises, pitches, spectra, silences, transitions, and durations.The articulatory form is the continuous representation of the gestures of the human sound-producing mechanism; it consists of the activities of the relevant muscles of the lungs, tongue, throat, larynx, lips and nose and their coordinations.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 74 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.